1 / 19

Structural Inference of Hierarchies in Networks

Structural Inference of Hierarchies in Networks. BY Yu Shuzhi 27, Mar 2014. Content. 1. Background 2 . Hierarchical Structures 3 . Random Graph Model of Hierarchical Organization 4. Consensus Hierarchies 5. Edge and Node Annotation 6. Prediction of Missing Interactions in Network

ingrid
Download Presentation

Structural Inference of Hierarchies in Networks

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. StructuralInferenceofHierarchiesinNetworks BYYuShuzhi 27,Mar2014

  2. Content • 1.Background • 2.HierarchicalStructures • 3.RandomGraph Model of Hierarchical Organization • 4. Consensus Hierarchies • 5. Edge and Node Annotation • 6. Prediction of Missing Interactions in Network • 7. Testing • 8. Work to do

  3. Background • Network and graph is a useful tool for analyzing complex systems. • Researchers try to develop new techniques and models for the analysis and interpretation of the network and graphs. • Hierarchy is an important property of real-world networks, since it can be observed in many networks.

  4. Background • Previously, Hierarchical Clustering algorithms are used to analyze the hierarchical structure. • Choose a similarity measure method • Compute similarity for each pair of vertices(n×n matrix) • Identify groups of vertices with high similarity • Agglomerative algorithms (iteratively merged) • Divisive algorithms (iteratively split)

  5. Background • Weakness of Traditional Hierarchical Clustering algorithms: • The algorithm only provides a single structure • It is unclear the result is unbiased.

  6. Hierarchical Structure • Definition of Hierarchical Structure: • It is one that divides naturally into groups and these groups themselves divide into subgroups, and so on until reaching the level of individual vertices. • Representations: • Dendrogram or Trees • Example of dendrogram:leavesaregraphverticesandinternalverticesrepresenthierarchicalrelationships

  7. RandomGraphModel • Assumption: • Theedgesofthegraphexistindependentlybutwithaprobabilitythatisnotidenticallydistributed.TheprobabilityisrepresentedasΘi. • HowtodetermineΘi: • Foradendrogram,usethemethodofmaximumlikelihoodtoestimate Θi. • Θi=E/(Li*Ri) • Ei:thenumberofedgesingraphthathavelowestcommonancestori(theinternalnode) • LiandRi:numberofleavesintheleft-andright-subtreerootedati. • Thelikelihoodforthedendrogramis: • LH(D, Θ)= Πi=1n-1(Θi)Ei(1– Θi)Li*Ri-Ei

  8. RandomGraphModel • Howtofindthedendrogramwiththemaximumlikelihood: • Itisdifficulttomaximizetheresultinglikelihood. • EmployaMarkovChainMonteCarlo(MCMC)method. • Thenumberofdendrogramswithnleavesissuper-exponential:(2n-3)!!.However,inpracticetheMCMCprocessworksrelativelyquicklyfornetworksuptoafewthousandvertices.

  9. RandomGraphModel • Markov Chain Monte Carlo sampling: • Let v denote the current state(adendrogram) of the Markov Chain. • EachinternalnodeIofthedendrogramisassociatedwiththreesubtress:twoareitschildrenandoneisitssibling.Therearethreeconfigurations. abcabcacb • Eachtimefortransition,chooseaninternalnoderandomlyandthenchooseoneofitstwoalternateconfigurationsuniformlyatrandom. For larger graphs, we can apply more dramaticallychangeofthestructure. • Weonlyacceptatransitionthatyieldsanincreaseinlikelihoodornochange:Lμ>=Lv;otherwise,acceptatransitionthatdecreasesthelikihoodwithprobabilityequaltotheratiooftherespectivestatelikelihoods: Lμ/Lv=elogLv-logLμ

  10. RandomGraphModel • After a while, the Markov Chain generates dendrogramsμat equilibrium with probabilities proportional to Lμ.

  11. ConsensusHierarchies • The idea is : • Instead of using one dendrogram to represent the hierarchical structure of the graph, we compute average features of the dendrogramsover the equilibrium distribution of models. • Method: • Takethecollectionofdendrogramsatequilibrium. • Derivesamajorityconsensusdendrogramcontainingonlythosehierarchicalfeaturesthathavemajorityweight. • Theweighthereisrepresentedbythelikelihoodofthedendrogram. • Result: • Theresultinghierarchicalstructuresisabettersummaryofthenetwork’sstructure. • Somecoarseningofthehierarchystructuresareremoved.

  12. RandomGraphModel • Examples: Originaldendrogramconsensusdendrogram

  13. NodeandEdgeAnnotation • Similartotheconceptofconsensus,wecanassignmajority-weightpropertiestonodesandedges. • Throughweightingeachdendrogramatequilibrium bylikelihood • For node, measure the average probability that a node belongs to its native group’ssubtree. • Foredge,measuretheaverageprobabilitythatanedgeexists. • Benefits: • Allowustoannotatethenetwork,highlightingthemostplausiblefeatures.

  14. NodeandEdgeAnnotation • Example: Annotatedversion: Linethicknessforedgesproportionaltotheiraverageprobabilityofexistance Shapeindicatesgroup Shadedproportionaltothesampledweightoftheirnativegroupaffiliation(lighter,higherprobability)

  15. PredictionofMissingInteractionsin Network • Hierarchical decomposition method: Find those highly possible connections but unconnected in real graph. These connections are probably missed. • Previous methods: • Assume that vertices are likely to be connected if • They have many common neighbors • There are short paths between them • They work well for strongly assortative networks, like citation and terrorist network. • Not good for disassortative network, like food webs.

  16. PredictionofMissingInteractionsin Network • Hierarchical decomposition method works well for both assortative and disassortative networks.

  17. Testing • Provided program: • fitHRG: • input a graph(edges list); • Output HierarchicalRandom Graph • ConsensusHRG: • input a dendrogram from fitHRG program • OutputHierarchicalRandom Graph • PredictHRG • Input a graph(edges list) • Output list of non-edges ranked by their model-averaged likelihood • Benchmark Test program provides: • Input a graph(edges list) • A list of nodes and their membership for the micro-communities • A list of nodes and their membership for the macro-communities

  18. Work to do • Figure out how to convert dendrogram into group list • Improve the algorithm and compare

  19. References • A. Clauset, C. Moore, and M.E.J. Newman. In E. M. Airoldi et al. (Eds.): ICML 2006 Ws, Lecture Notes in Computer Science4503, 1 - 13. Springer-Verlag, Berlin Heidelberg (2007). • A. Clauset, C. Moore, and M.E.J. Newman. Nature453, 98 - 101 (2008)

More Related